Prenatal diagnosis of fetal skeletal dysplasia using targeted next-generation sequencing: an analysis of 30 cases

The present study comprised of 30 cases of fetuses diagnosed with skeletal abnormalities via ultrasound at the Obstetrics and Gynecology Hospital of Capital Medical University, Beijing, China between January 2014 and June 2017. Among these 30 cases, 15 cases were fetal skeletal malformations, while 15 cases were systemic skeletal dysplasias, which were characterized as short limb deformities. For the diagnosis of short limb deformity, two criteria must be simultaneously satisfied. First, the long bones of the extremities (i.e. the femur and humerus) must be shorter than the 5th percentile for fetuses of the same gestational age. In addition, the femur length (FL) to abdominal circumference (AC) ratio must be ≤1.6. Second, there must be ultrasonographic manifestations of abnormal bone morphology, such as long bone bending, angulation, fractures, “telephone receiver-shaped” changes, thoracic dysplasia, or changes in bone mineral density [10]. The present study was approved by the ethics committee of our hospital, and all parents of the fetuses provided a signed informed consent prior to prenatal diagnosis and sample collection.

After routine disinfection, fetal umbilical cord blood puncture was performed, and 2 mL of blood was drawn from the umbilical vein into a vacuum blood collection tube containing ethylenediaminetetraacetic acid (EDTA). If no fetal blood sample could be obtained, two pieces of fetal muscle tissue (3 × 3 cm) containing the skin were removed after abortion. For all cases, 5 mL of venous blood was collected from both parents in vacuum blood collection tubes containing EDTA. All specimens were treated and stored at − 80 °C until use. Both the sources of the samples and methods of blood sampling were approved by the Institutional Ethics Committee of the Obstetrics and Gynecology Hospital of Capital Medical University. The informed consent forms were completed by the parents.

Fetal amniotic fluid or umbilical cord blood samples were taken for chromosome G band karyotype analysis.

Umbilical cord blood punctures were performed on fetuses with abnormal skeletal abnormalities. Complete genomic DNA was extracted using a commercial DNA extraction kit (Puregene; Qiagen, Hilden, Germany) from either the umbilical cord blood, or muscle tissue samples. WGS with 20–30× coverage, combined with bioinformatics analysis, accurately localize microdeletions and microduplications ≥100 kb long. If ≥100 kb disease-related microdeletions are detected in fetal tissues, these results were verified in the parent samples using Sanger sequencing.

If neither chromosomal abnormalities, nor disease-related microdeletions/microduplications were detected, the protein-coding regions and adjacent regions of 30 bp of known genes related to congenital skeletal anomalies were deeply sequenced using the targeted gene sequencing method. A special capture array (BGI, China) was used to detect the variants of the 363 genes involved in congenital skeletal anomalies. The overall sequencing coverage of the target regions was 97.81% for 20× depth of coverage in each of the chromosomes. After filtration by SAMtools (version 1.4), the sequencing data were mapped to the human genome (NCBI37/hg19) using the Burrows Wheeler Aligner software, and single nucleotide variants were identified using the SOAPsnp software (version 2.0). Then, the biological information was compared to databases, including ExAC, dbSNP, HapMap, 1,000 Genomes Asian, ESP6500, Cosmic, and HGMD. All variants were classified according to American College of Medical Genetics and Genomics recommended standards. After matching the inherent patterns of the disease, the pathogenicity of these loci were further determined by assessing the clinical symptoms and genetic data.

If pathogenic mutations were detected in the fetal samples, these mutations were further verified in the parents by Sanger sequencing.

Among the 30 cases in the present study, 15 fetuses had local deformities (including varus deformities, finger/toe deformities, missing fingers/toes, and/or absence of upper/lower limbs), while 15 fetuses had systemic general skeletal dysplasia characterized by short limb deformities. All deformities were confirmed by postpartum clinical and pathological analysis (Table 1).

All 30 cases of fetal skeletal dysplasia were tested for chromosomal abnormalities. Trisomy 18 was detected in two cases of fetal local skeletal malformation (cases 1–2, Table 1). However, neither chromosomal abnormalities, nor pathological microdeletions/microduplications related to skeletal dysplasia were identified in the remaining 28 cases.

Among the 13 cases of fetal local skeletal malformation, in which no chromosomal abnormalities were identified, seven cases (cases 9–15) had no pathogenic mutations, while six cases (cases 3–8) had mutations in known genes associated with bone diseases (Table 1). In three of these cases (cases 3–5), the mutations were of unknown clinical significance, while in the remaining three cases (cases 6–8), the mutations were considered as causative.

In case 22, a heterozygous mutation in the emopamil binding protein (EBP) gene (NM_006579.2) was detected (C.440G > A, p.Arg147His). In case 30, a clinically significant heterozygous mutation in the filamin B (FLNB) gene was detected (c.475A > C, p.Thr159Pro). For both cases, the parents were not the carriers, suggesting that these were new fetal mutations. The mother of case 22 has subsequently given birth to a healthy baby boy (Table 2).

Different types of chromosomal abnormalities complicate the wide, diverse variety of skeletal abnormalities [16]. For example, fetuses carrying trisomy 13, trisomy 18, or even trisomy 21 (Down’s syndrome) may have abnormal skeletal development. In the present study, two fetuses with trisomy 18 (cases 1 and 2) exhibited chromosomal abnormalities in local bone lesions, suggesting that the screening for chromosomal abnormalities remains vital when skeletal lesions are observed (Fig. 1).

The WGS revealed no significant copy number variations in the 13 cases of local fetal skeletal abnormalities (cases 3–15, Table 1). In addition, no mutations of skeletal pathogenic genes were identified in the seven cases (cases 9–15, Table 2). In contrast to osteodystrophy, local skeletal abnormalities are regulated by other factors in addition to hereditary genes. However, these additional factors require further study.

In the six fetuses with skeletal lesions (cases 3–8), skeletal gene mutations were detected with targeted gene sequencing (Table 2). In three of these cases (cases 3–5), the fetuses carried heterozygous mutations of unclear clinical significance. Both case 3 (femoral absence, micrognathia, and cleft lip and palate) and case 4 (spinal fracture, ventral fissure, introversion, and left foot polydactyly) carried clinically unexplained heterozygous mutations. The fetus in case 3 carried TBX4-induced small zygomatic complexes, while the fetus in case 4 carried TNNT3-related distal type 2B. Both of these are autosomal dominant genetic diseases. However, the Sanger sequencing suggested that one of the healthy parents carried the mutation. Hence, these two mutations could not explain the reason of case 3 and 4.

In case 5, the fetus carried a clinically unexplained heterozygous mutation in the SOX9 gene. SOX9-related trunk dysplasia is an autosomal dominant genetic disease, and in such diseases, heterozygous mutations can induce a disease phenotype. In this case, the Sanger sequencing indicated that neither of the parents were mutation carriers. Hence, this mutation was considered de novo and likely pathologic. Notably, the mother became naturally pregnant again and gave birth to a healthy baby boy.

In the remaining three cases (cases 6–8), known disease-causing mutations were detected in the fetuses. The available information about these mutations was useful in fetal disease diagnosis and in the subsequent genetic counseling for the parents.

In case 6 (fetal cleft lip and palate, right kidney cystic hypoplasia, bilateral foot fissure, lateral toe, and abnormal hand development), the fetus carried a mutation in the TP63 gene, which was correlated to ectrodactyly, ectoderm dysplasia, and cleft lip/palate syndrome-3 (EEC3). EEC3 is an autosomal dominant genetic disease, and in such diseases, heterozygous mutations can cause the disease. Alves et al. [17], Hydern et al. [18] and Clements et al. [19] all reported several mutations of the TP63 gene in families of EEC3. The Sanger sequencing revealed negative results for both parents of case 6, and this mutation was thereby considered novel in the fetus. The subsequent genetic counseling advised the parents to continue to conceive naturally.

In case 7 (angular femoral deformity and ventricular septal defect), two compound heterozygous causative mutations in the POR gene were detected: c.1370G > A (p.Arg457His) and c.744C > G (p.Tyr248*), which were associated with the Antley-Bixler syndrome. The Antley-Bixler syndrome is an autosomal recessive disease, and this type of homozygous or compound heterozygous mutation can induce a disease phenotype. The mutation c.1370G > A (p.Arg457His) has been detected in patients diagnosed with Antley-Bixler syndrome [20, 21]. The other mutations have also been reported in recent years [22, 23]. The mutation c.744C > G (p.Tyr248*) is a nonsense mutation that prematurely terminates the encoding of the POR protein at amino acid 248 (the unmutated protein is 680 amino acids long). Although the specific mutation identified in the present study has not yet been reported in literature, previously reported nonsense mutations after position 248 were deleterious. POR gene mutations c.1370G > A and c.744C > G have been considered disease-causing in their compound heterozygous forms. In the present study, the Sanger sequencing verified that both parents were carriers of this disease-causing mutation. The parents chose to conceive naturally at 6 months after pregnancy termination. However, the pregnancy was terminated at the 15th week of the second pregnancy, because the ultrasound examination revealed bilateral femoral angulation again in the fetus. Compound heterozygous POR gene mutations c.1370G > A and c.744C > G were verified on the analysis of the aborted tissues. The parents conceived naturally for the third time. The ultrasound results were normal, and the amniotic fluid analysis indicated that the POR gene mutation was absent. The mother eventually delivered a healthy baby girl.

In the present study, no chromosomal abnormalities were identified in any of the cases of systemic skeletal dysplasia that manifested in short limb deformities (Table 1). The targeted gene sequencing identified several disease-causative mutations in known genes related to the disease (Table 2).

Collagen is an indispensable component of bone tissue found in the extracellular matrix. Mutations in the collagen gene may lead to insufficient collagen production. There are substantial differences in the severity of skeletal abnormalities caused by different types of collagen mutations. Mutations in collagen genes were detected in cases 16–21 in the present study. Three fetuses (cases 16–18) were diagnosed with osteogenesis imperfecta based on prenatal ultrasound, gross postnatal pathology, and X-ray examination (Fig. 2). Heterozygous mutations in COL2A1 were found in the other two fetuses (cases 19 and 20) diagnosed with achondrogenesis type II.

In case 16, a heterozygous causative mutation (c.1678G > A, p.Gly560Ser) in COL1A1 was detected in the fetus. In cases 17 and 18, two known pathogenic mutations were detected in the COL1A2 gene: c.1774G > A (p.Gly592Ser) and c.1072G > A (p.Gly358Ser). These mutations have been reported to be pathogenic mutations associated with osteogenesis imperfecta [24‐28]. However, none of these mutations were detected in the parents, suggesting that these mutations are novel fetal mutations.

Approximately 90% of osteogenesis imperfecta cases are due to causative variants in the COL1A1 and COL1A2 genes, which result in abnormal collagen I fibrils formation, while the remaining 10% of cases are associated with recessive variants of known or yet to be discovered genes [24, 25].

Heterozygous mutations in COL2A1 were found in two fetuses (cases 19 and 20) diagnosed with achondrogenesis type II (Fig. 3). Mutations in COL2A1 disrupt the Gly-XY motif necessary for the formation of a triple helix structure, resulting in type II collagen over-modification, cellular retention and decreased secretion [29, 30]. All these collagen-related mutations were new in the fetuses, and the parents were not mutation carriers. The parents were advised to continue natural conception.

Case 21 was eventually diagnosed with fibrocartilage hyperplasia type II, which was caused by mutations in the COL11A2 gene. The two COL11A2 mutations identified in the present study have only been previously identified once [31]. These mutations revealed autosomal recessive/dominant inheritance. The Sanger sequencing verified that these phenotypically normal parents were mutation carriers. These parents were trying to conceive again with PGD.

In the present study, seven fetuses with achondroplasia (cases 23–29) were further examined with a combination of ultrasound, postnatal gross pathology, and X-ray. Based on these examinations, six of the fetuses were diagnosed with clinically fatal cartilage hypoplasia. The subsequent genetic analysis confirmed the fatal cartilage hypoplasia type I in six fetuses. Cases 23–26 carried an identical mutation (c.742C > T, p.Arg248Cys), which is a common pathogenic mutation associated with lethal achondroplasia (Fig. 4). In 2015, Barkova et al. [2] reported that eight of 20 patients with lethal dysplasia type I carried the c.742C > T mutation in FGFR3. Similar findings were reported in 2001 by Chen et al. [32] and in 2014 by Cho et al. [33]. The mutation c.1124A > G (p.Tyr375Cys) in the FGFR3 gene is also a common pathogenic mutation that causes lethal dwarfism type I. The study conducted by Rousseau et al. [34] revealed that eight of 26 (30.7%) cases of fatal dwarfism type I carried the c.1124A > G (i.e. c.1118A > G in the article) mutation in FGFR3. However, Xue et al. [35] reported that the frequency of these FGFR3 mutations was 23.7% (41 of 173 cases) in cases of fatal dwarfism type I. Another FGFR3 mutation, c.2426G > C (p.X809S,101), was identified in the present study, which has not previously been reported. This was a missense mutation that led to the false extension of protein translation. In case 27, the fetus was diagnosed with achondroplasia. Furthermore, 99% of all achondroplasia cases are caused by mutations in the FGFR3 gene, and c.1144G > A (p.Gly382Arg) is the most common pathogenic mutation. The missense mutation c.1144G > A (p.Gly382Arg) is identical to c.1138G > A (p.Gly380Arg) (different transcripts). In 1995, the study conducted by Bellus et al. [36] revealed that 187 of 193 (96.9%) cases of achondroplasia are caused by the mutation c.1138G > A. For case 27, since the parents did not carry the mutation, it was considered a new mutation in the fetus. After genetic counseling, the parents were advised to continue to conceive naturally.

The mutation c.475A > C (p.Thr159Pro) in the FLNB gene carried by the fetus in case 30 is a missense mutation, which changes the amino acid at position 159 from threonine to proline. FLNB-related osteogenesis imperfecta type I/Larsen syndrome is an autosomal dominant disease. The sequence verification confirmed that neither of the parents carried the mutation, indicating that this was a novel mutation in the fetus. Therefore, the parents were advised to continue to conceive naturally.

In case 22, ultrasound revealed the abnormal development of long bones of the limbs (the length of the long bones was less than 1%), thick metaphysis in the right lower limb, irregular vertebral arrangement, and a narrow and small thorax in 24 weeks of gestation. The fetus carried a heterozygous mutation in the EBP gene (C.440G > A, p.Arg147His). Herman et al. investigated the mutations in 26 female patients with suspected CDPX2. Among these 26 patients, 22 had EBP mutations. Among these 22 mutations, 13 mutations were de novo [37]. The EBP gene was located on the short arm of the X chromosome (Xp11.22-p11.23), and the mutations in this gene can lead to the accumulation of 8-dehydrocholesterol and 8(9)-cholesterol in plasma, the skin and other tissues, resulting in a wide range of abnormalities [37, 38]. In the present study, a heterozygous EBP mutation (a known causative mutation; c.440G > A [p.Arg147His]) was the cause of CDPX2, and is a missense mutation (arginine to histidine). The Sanger test revealed that the patient’s parents were non-carriers, indicating that the mutation of the fetus occurred de novo. Whittock et al., Has et al. and Cañueto et al. have previously reported cases related to this mutation site. [39‐41]. However, neither of the parents carried the mutation. In the present study, the 27-week-old female fetus presented with markedly short bones, ankle joint contracture, markedly asymmetric short lower limbs, and a flat face and nose bridge. The severity of the phenotype was considered to be related to X chromosome inactivation, which is also known as lyonization.

However, although targeted exome capture and sequencing have shown great advantages in disease gene identification and molecular diagnosis, some problems still needs to be immediately resolved. Fox example, exome sequencing focuses on the sequencing of exon regions. Thus, from the genome level, the information obtained was obviously incomplete. Furthermore, information for promoter regions, enhancer regions and microRNA coding regions were certainly missed. Second, a large amount of data was obtained after exome sequencing. The best method to perform an in-depth and accurate analysis of these data is the largest challenge faced at present by researchers worldwide. The deep mining of data needs to start from many aspects and perspectives, including studies at the transcription level, bioinformatics analysis, and functional genomics studies.